Abstract 1104: Identifying pancreatic cancer subtypes by a novel meta-learning model

Abstract Due to the difficulty of both early diagnosis and treatment, pancreatic cancer (PC) is the third leading cause of cancer related deaths in the United States and is projected to become the second by 2030. As a heterogeneous disease, PC can be divided into four main subtypes, with each having differing molecular pathways, survival rates, and responses to therapy. Traditional wet lab approaches to identify PC subtypes, such as histopathological studies or molecular profiling, are costly, laborious, and time-consuming. Previous studies indicate that predicting PC subtypes based on transcriptomic profiling is a feasible and promising method. However, these methods are unable to capture the full complexity of the data and may lead to biased and unstable results. To this end, we present a novel meta-learning algorithm for PC subtype identification based on RNA-seq data only. Specifically, after some basic gene filtering and preprocessing, the transcriptome-based feature vectors were classified by multiple base machine learning (ML) classifiers like logistic regression, linear support vector machine (SVM), and radial basis function (RBF) SVM, whose outputs were then combined with the initial filtered feature vectors to constitute a new set of ensemble feature vectors for a meta-learning model. Our meta-learning model could learn and leverage the diversity of different base classifiers to boost the prediction performance beyond any single ML model. Results based on 100 times leave-one-out cross validation tests on a dataset with 96 PC patients demonstrated that our proposed method could achieve an average AUC of 0.99, which was significantly better than existing state-of-the-art methods. In addition, our meta-learning model remarkably outperformed each individual base classifier, demonstrating that our model could combine diverse results from multiple base classifiers to boost the ensemble performance. We expect that our method will provide a complementary way to characterize different PC subtypes, bringing positive impacts on downstream pancreatic cancer risk stratification and tailored treatment design. Additionally, we believe our meta-learning model can also be extensible to identify subtypes of other types of cancer. Citation Format: Nick Boos Peterson, Mengtao Sun, Jieqiong Wang, Shibiao Wan. Identifying pancreatic cancer subtypes by a novel meta-learning model [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2025; Part 1 (Regular Abstracts); 2025 Apr 25-30; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2025;85(8_Suppl_1):Abstract nr 1104.